summaryrefslogtreecommitdiffstats
path: root/fs/ocfs2/dlm/dlmmaster.c
blob: 856012b4fa4954534ac6d4471d2ebee25b9bdae9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * dlmmod.c
 *
 * standalone DLM module
 *
 * Copyright (C) 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 *
 */


#include <linux/module.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/utsname.h>
#include <linux/init.h>
#include <linux/sysctl.h>
#include <linux/random.h>
#include <linux/blkdev.h>
#include <linux/socket.h>
#include <linux/inet.h>
#include <linux/spinlock.h>
#include <linux/delay.h>


#include "cluster/heartbeat.h"
#include "cluster/nodemanager.h"
#include "cluster/tcp.h"

#include "dlmapi.h"
#include "dlmcommon.h"
#include "dlmdomain.h"

#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_MASTER)
#include "cluster/masklog.h"

enum dlm_mle_type {
	DLM_MLE_BLOCK,
	DLM_MLE_MASTER,
	DLM_MLE_MIGRATION
};

struct dlm_lock_name
{
	u8 len;
	u8 name[DLM_LOCKID_NAME_MAX];
};

struct dlm_master_list_entry
{
	struct list_head list;
	struct list_head hb_events;
	struct dlm_ctxt *dlm;
	spinlock_t spinlock;
	wait_queue_head_t wq;
	atomic_t woken;
	struct kref mle_refs;
	int inuse;
	unsigned long maybe_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
	unsigned long vote_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
	unsigned long response_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
	unsigned long node_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
	u8 master;
	u8 new_master;
	enum dlm_mle_type type;
	struct o2hb_callback_func mle_hb_up;
	struct o2hb_callback_func mle_hb_down;
	union {
		struct dlm_lock_resource *res;
		struct dlm_lock_name name;
	} u;
};

static void dlm_mle_node_down(struct dlm_ctxt *dlm,
			      struct dlm_master_list_entry *mle,
			      struct o2nm_node *node,
			      int idx);
static void dlm_mle_node_up(struct dlm_ctxt *dlm,
			    struct dlm_master_list_entry *mle,
			    struct o2nm_node *node,
			    int idx);

static void dlm_assert_master_worker(struct dlm_work_item *item, void *data);
static int dlm_do_assert_master(struct dlm_ctxt *dlm, const char *lockname,
				unsigned int namelen, void *nodemap,
				u32 flags);

static inline int dlm_mle_equal(struct dlm_ctxt *dlm,
				struct dlm_master_list_entry *mle,
				const char *name,
				unsigned int namelen)
{
	struct dlm_lock_resource *res;

	if (dlm != mle->dlm)
		return 0;

	if (mle->type == DLM_MLE_BLOCK ||
	    mle->type == DLM_MLE_MIGRATION) {
		if (namelen != mle->u.name.len ||
    	    	    memcmp(name, mle->u.name.name, namelen)!=0)
			return 0;
	} else {
		res = mle->u.res;
		if (namelen != res->lockname.len ||
		    memcmp(res->lockname.name, name, namelen) != 0)
			return 0;
	}
	return 1;
}

#define dlm_print_nodemap(m)  _dlm_print_nodemap(m,#m)
static void _dlm_print_nodemap(unsigned long *map, const char *mapname)
{
	int i;
	printk("%s=[ ", mapname);
	for (i=0; i<O2NM_MAX_NODES; i++)
		if (test_bit(i, map))
			printk("%d ", i);
	printk("]");
}

static void dlm_print_one_mle(struct dlm_master_list_entry *mle)
{
	int refs;
	char *type;
	char attached;
	u8 master;
	unsigned int namelen;
	const char *name;
	struct kref *k;
	unsigned long *maybe = mle->maybe_map,
		      *vote = mle->vote_map,
		      *resp = mle->response_map,
		      *node = mle->node_map;

	k = &mle->mle_refs;
	if (mle->type == DLM_MLE_BLOCK)
		type = "BLK";
	else if (mle->type == DLM_MLE_MASTER)
		type = "MAS";
	else
		type = "MIG";
	refs = atomic_read(&k->refcount);
	master = mle->master;
	attached = (list_empty(&mle->hb_events) ? 'N' : 'Y');

	if (mle->type != DLM_MLE_MASTER) {
		namelen = mle->u.name.len;
		name = mle->u.name.name;
	} else {
		namelen = mle->u.res->lockname.len;
		name = mle->u.res->lockname.name;
	}

	mlog(ML_NOTICE, "%.*s: %3s refs=%3d mas=%3u new=%3u evt=%c inuse=%d ",
		  namelen, name, type, refs, master, mle->new_master, attached,
		  mle->inuse);
	dlm_print_nodemap(maybe);
	printk(", ");
	dlm_print_nodemap(vote);
	printk(", ");
	dlm_print_nodemap(resp);
	printk(", ");
	dlm_print_nodemap(node);
	printk(", ");
	printk("\n");
}

#if 0
/* Code here is included but defined out as it aids debugging */

static void dlm_dump_mles(struct dlm_ctxt *dlm)
{
	struct dlm_master_list_entry *mle;
	struct list_head *iter;
	
	mlog(ML_NOTICE, "dumping all mles for domain %s:\n", dlm->name);
	spin_lock(&dlm->master_lock);
	list_for_each(iter, &dlm->master_list) {
		mle = list_entry(iter, struct dlm_master_list_entry, list);
		dlm_print_one_mle(mle);
	}
	spin_unlock(&dlm->master_lock);
}

int dlm_dump_all_mles(const char __user *data, unsigned int len)
{
	struct list_head *iter;
	struct dlm_ctxt *dlm;

	spin_lock(&dlm_domain_lock);
	list_for_each(iter, &dlm_domains) {
		dlm = list_entry (iter, struct dlm_ctxt, list);
		mlog(ML_NOTICE, "found dlm: %p, name=%s\n", dlm, dlm->name);
		dlm_dump_mles(dlm);
	}
	spin_unlock(&dlm_domain_lock);
	return len;
}
EXPORT_SYMBOL_GPL(dlm_dump_all_mles);

#endif  /*  0  */


static struct kmem_cache *dlm_mle_cache = NULL;


static void dlm_mle_release(struct kref *kref);
static void dlm_init_mle(struct dlm_master_list_entry *mle,
			enum dlm_mle_type type,
			struct dlm_ctxt *dlm,
			struct dlm_lock_resource *res,
			const char *name,
			unsigned int namelen);
static void dlm_put_mle(struct dlm_master_list_entry *mle);
static void __dlm_put_mle(struct dlm_master_list_entry *mle);
static int dlm_find_mle(struct dlm_ctxt *dlm,
			struct dlm_master_list_entry **mle,
			char *name, unsigned int namelen);

static int dlm_do_master_request(struct dlm_master_list_entry *mle, int to);


static int dlm_wait_for_lock_mastery(struct dlm_ctxt *dlm,
				     struct dlm_lock_resource *res,
				     struct dlm_master_list_entry *mle,
				     int *blocked);
static int dlm_restart_lock_mastery(struct dlm_ctxt *dlm,
				    struct dlm_lock_resource *res,
				    struct dlm_master_list_entry *mle,
				    int blocked);
static int dlm_add_migration_mle(struct dlm_ctxt *dlm,
				 struct dlm_lock_resource *res,
				 struct dlm_master_list_entry *mle,
				 struct dlm_master_list_entry **oldmle,
				 const char *name, unsigned int namelen,
				 u8 new_master, u8 master);

static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
				    struct dlm_lock_resource *res);
static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
				      struct dlm_lock_resource *res);
static int dlm_mark_lockres_migrating(struct dlm_ctxt *dlm,
				       struct dlm_lock_resource *res,
				       u8 target);
static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
				       struct dlm_lock_resource *res);


int dlm_is_host_down(int errno)
{
	switch (errno) {
		case -EBADF:
		case -ECONNREFUSED:
		case -ENOTCONN:
		case -ECONNRESET:
		case -EPIPE:
		case -EHOSTDOWN:
		case -EHOSTUNREACH:
		case -ETIMEDOUT:
		case -ECONNABORTED:
		case -ENETDOWN:
		case -ENETUNREACH:
		case -ENETRESET:
		case -ESHUTDOWN:
		case -ENOPROTOOPT:
		case -EINVAL:   /* if returned from our tcp code,
				   this means there is no socket */
			return 1;
	}
	return 0;
}


/*
 * MASTER LIST FUNCTIONS
 */


/*
 * regarding master list entries and heartbeat callbacks:
 *
 * in order to avoid sleeping and allocation that occurs in
 * heartbeat, master list entries are simply attached to the
 * dlm's established heartbeat callbacks.  the mle is attached
 * when it is created, and since the dlm->spinlock is held at
 * that time, any heartbeat event will be properly discovered
 * by the mle.  the mle needs to be detached from the
 * dlm->mle_hb_events list as soon as heartbeat events are no
 * longer useful to the mle, and before the mle is freed.
 *
 * as a general rule, heartbeat events are no longer needed by
 * the mle once an "answer" regarding the lock master has been
 * received.
 */
static inline void __dlm_mle_attach_hb_events(struct dlm_ctxt *dlm,
					      struct dlm_master_list_entry *mle)
{
	assert_spin_locked(&dlm->spinlock);

	list_add_tail(&mle->hb_events, &dlm->mle_hb_events);
}


static inline void __dlm_mle_detach_hb_events(struct dlm_ctxt *dlm,
					      struct dlm_master_list_entry *mle)
{
	if (!list_empty(&mle->hb_events))
		list_del_init(&mle->hb_events);
}


static inline void dlm_mle_detach_hb_events(struct dlm_ctxt *dlm,
					    struct dlm_master_list_entry *mle)
{
	spin_lock(&dlm->spinlock);
	__dlm_mle_detach_hb_events(dlm, mle);
	spin_unlock(&dlm->spinlock);
}

static void dlm_get_mle_inuse(struct dlm_master_list_entry *mle)
{
	struct dlm_ctxt *dlm;
	dlm = mle->dlm;

	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&dlm->master_lock);
	mle->inuse++;
	kref_get(&mle->mle_refs);
}

static void dlm_put_mle_inuse(struct dlm_master_list_entry *mle)
{
	struct dlm_ctxt *dlm;
	dlm = mle->dlm;

	spin_lock(&dlm->spinlock);
	spin_lock(&dlm->master_lock);
	mle->inuse--;
	__dlm_put_mle(mle);
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);

}

/* remove from list and free */
static void __dlm_put_mle(struct dlm_master_list_entry *mle)
{
	struct dlm_ctxt *dlm;
	dlm = mle->dlm;

	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&dlm->master_lock);
	if (!atomic_read(&mle->mle_refs.refcount)) {
		/* this may or may not crash, but who cares.
		 * it's a BUG. */
		mlog(ML_ERROR, "bad mle: %p\n", mle);
		dlm_print_one_mle(mle);
		BUG();
	} else
		kref_put(&mle->mle_refs, dlm_mle_release);
}


/* must not have any spinlocks coming in */
static void dlm_put_mle(struct dlm_master_list_entry *mle)
{
	struct dlm_ctxt *dlm;
	dlm = mle->dlm;

	spin_lock(&dlm->spinlock);
	spin_lock(&dlm->master_lock);
	__dlm_put_mle(mle);
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);
}

static inline void dlm_get_mle(struct dlm_master_list_entry *mle)
{
	kref_get(&mle->mle_refs);
}

static void dlm_init_mle(struct dlm_master_list_entry *mle,
			enum dlm_mle_type type,
			struct dlm_ctxt *dlm,
			struct dlm_lock_resource *res,
			const char *name,
			unsigned int namelen)
{
	assert_spin_locked(&dlm->spinlock);

	mle->dlm = dlm;
	mle->type = type;
	INIT_LIST_HEAD(&mle->list);
	INIT_LIST_HEAD(&mle->hb_events);
	memset(mle->maybe_map, 0, sizeof(mle->maybe_map));
	spin_lock_init(&mle->spinlock);
	init_waitqueue_head(&mle->wq);
	atomic_set(&mle->woken, 0);
	kref_init(&mle->mle_refs);
	memset(mle->response_map, 0, sizeof(mle->response_map));
	mle->master = O2NM_MAX_NODES;
	mle->new_master = O2NM_MAX_NODES;
	mle->inuse = 0;

	if (mle->type == DLM_MLE_MASTER) {
		BUG_ON(!res);
		mle->u.res = res;
	} else if (mle->type == DLM_MLE_BLOCK) {
		BUG_ON(!name);
		memcpy(mle->u.name.name, name, namelen);
		mle->u.name.len = namelen;
	} else /* DLM_MLE_MIGRATION */ {
		BUG_ON(!name);
		memcpy(mle->u.name.name, name, namelen);
		mle->u.name.len = namelen;
	}

	/* copy off the node_map and register hb callbacks on our copy */
	memcpy(mle->node_map, dlm->domain_map, sizeof(mle->node_map));
	memcpy(mle->vote_map, dlm->domain_map, sizeof(mle->vote_map));
	clear_bit(dlm->node_num, mle->vote_map);
	clear_bit(dlm->node_num, mle->node_map);

	/* attach the mle to the domain node up/down events */
	__dlm_mle_attach_hb_events(dlm, mle);
}


/* returns 1 if found, 0 if not */
static int dlm_find_mle(struct dlm_ctxt *dlm,
			struct dlm_master_list_entry **mle,
			char *name, unsigned int namelen)
{
	struct dlm_master_list_entry *tmpmle;
	struct list_head *iter;

	assert_spin_locked(&dlm->master_lock);

	list_for_each(iter, &dlm->master_list) {
		tmpmle = list_entry(iter, struct dlm_master_list_entry, list);
		if (!dlm_mle_equal(dlm, tmpmle, name, namelen))
			continue;
		dlm_get_mle(tmpmle);
		*mle = tmpmle;
		return 1;
	}
	return 0;
}

void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up)
{
	struct dlm_master_list_entry *mle;
	struct list_head *iter;

	assert_spin_locked(&dlm->spinlock);
	
	list_for_each(iter, &dlm->mle_hb_events) {
		mle = list_entry(iter, struct dlm_master_list_entry, 
				 hb_events);
		if (node_up)
			dlm_mle_node_up(dlm, mle, NULL, idx);
		else
			dlm_mle_node_down(dlm, mle, NULL, idx);
	}
}

static void dlm_mle_node_down(struct dlm_ctxt *dlm,
			      struct dlm_master_list_entry *mle,
			      struct o2nm_node *node, int idx)
{
	spin_lock(&mle->spinlock);

	if (!test_bit(idx, mle->node_map))
		mlog(0, "node %u already removed from nodemap!\n", idx);
	else
		clear_bit(idx, mle->node_map);

	spin_unlock(&mle->spinlock);
}

static void dlm_mle_node_up(struct dlm_ctxt *dlm,
			    struct dlm_master_list_entry *mle,
			    struct o2nm_node *node, int idx)
{
	spin_lock(&mle->spinlock);

	if (test_bit(idx, mle->node_map))
		mlog(0, "node %u already in node map!\n", idx);
	else
		set_bit(idx, mle->node_map);

	spin_unlock(&mle->spinlock);
}


int dlm_init_mle_cache(void)
{
	dlm_mle_cache = kmem_cache_create("dlm_mle_cache",
					  sizeof(struct dlm_master_list_entry),
					  0, SLAB_HWCACHE_ALIGN,
					  NULL, NULL);
	if (dlm_mle_cache == NULL)
		return -ENOMEM;
	return 0;
}

void dlm_destroy_mle_cache(void)
{
	if (dlm_mle_cache)
		kmem_cache_destroy(dlm_mle_cache);
}

static void dlm_mle_release(struct kref *kref)
{
	struct dlm_master_list_entry *mle;
	struct dlm_ctxt *dlm;

	mlog_entry_void();

	mle = container_of(kref, struct dlm_master_list_entry, mle_refs);
	dlm = mle->dlm;

	if (mle->type != DLM_MLE_MASTER) {
		mlog(0, "calling mle_release for %.*s, type %d\n",
		     mle->u.name.len, mle->u.name.name, mle->type);
	} else {
		mlog(0, "calling mle_release for %.*s, type %d\n",
		     mle->u.res->lockname.len,
		     mle->u.res->lockname.name, mle->type);
	}
	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&dlm->master_lock);

	/* remove from list if not already */
	if (!list_empty(&mle->list))
		list_del_init(&mle->list);

	/* detach the mle from the domain node up/down events */
	__dlm_mle_detach_hb_events(dlm, mle);

	/* NOTE: kfree under spinlock here.
	 * if this is bad, we can move this to a freelist. */
	kmem_cache_free(dlm_mle_cache, mle);
}


/*
 * LOCK RESOURCE FUNCTIONS
 */

static void dlm_set_lockres_owner(struct dlm_ctxt *dlm,
				  struct dlm_lock_resource *res,
				  u8 owner)
{
	assert_spin_locked(&res->spinlock);

	mlog_entry("%.*s, %u\n", res->lockname.len, res->lockname.name, owner);

	if (owner == dlm->node_num)
		atomic_inc(&dlm->local_resources);
	else if (owner == DLM_LOCK_RES_OWNER_UNKNOWN)
		atomic_inc(&dlm->unknown_resources);
	else
		atomic_inc(&dlm->remote_resources);

	res->owner = owner;
}

void dlm_change_lockres_owner(struct dlm_ctxt *dlm,
			      struct dlm_lock_resource *res, u8 owner)
{
	assert_spin_locked(&res->spinlock);

	if (owner == res->owner)
		return;

	if (res->owner == dlm->node_num)
		atomic_dec(&dlm->local_resources);
	else if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN)
		atomic_dec(&dlm->unknown_resources);
	else
		atomic_dec(&dlm->remote_resources);

	dlm_set_lockres_owner(dlm, res, owner);
}


static void dlm_lockres_release(struct kref *kref)
{
	struct dlm_lock_resource *res;

	res = container_of(kref, struct dlm_lock_resource, refs);

	/* This should not happen -- all lockres' have a name
	 * associated with them at init time. */
	BUG_ON(!res->lockname.name);

	mlog(0, "destroying lockres %.*s\n", res->lockname.len,
	     res->lockname.name);

	if (!hlist_unhashed(&res->hash_node) ||
	    !list_empty(&res->granted) ||
	    !list_empty(&res->converting) ||
	    !list_empty(&res->blocked) ||
	    !list_empty(&res->dirty) ||
	    !list_empty(&res->recovering) ||
	    !list_empty(&res->purge)) {
		mlog(ML_ERROR,
		     "Going to BUG for resource %.*s."
		     "  We're on a list! [%c%c%c%c%c%c%c]\n",
		     res->lockname.len, res->lockname.name,
		     !hlist_unhashed(&res->hash_node) ? 'H' : ' ',
		     !list_empty(&res->granted) ? 'G' : ' ',
		     !list_empty(&res->converting) ? 'C' : ' ',
		     !list_empty(&res->blocked) ? 'B' : ' ',
		     !list_empty(&res->dirty) ? 'D' : ' ',
		     !list_empty(&res->recovering) ? 'R' : ' ',
		     !list_empty(&res->purge) ? 'P' : ' ');

		dlm_print_one_lock_resource(res);
	}

	/* By the time we're ready to blow this guy away, we shouldn't
	 * be on any lists. */
	BUG_ON(!hlist_unhashed(&res->hash_node));
	BUG_ON(!list_empty(&res->granted));
	BUG_ON(!list_empty(&res->converting));
	BUG_ON(!list_empty(&res->blocked));
	BUG_ON(!list_empty(&res->dirty));
	BUG_ON(!list_empty(&res->recovering));
	BUG_ON(!list_empty(&res->purge));

	kfree(res->lockname.name);

	kfree(res);
}

void dlm_lockres_put(struct dlm_lock_resource *res)
{
	kref_put(&res->refs, dlm_lockres_release);
}

static void dlm_init_lockres(struct dlm_ctxt *dlm,
			     struct dlm_lock_resource *res,
			     const char *name, unsigned int namelen)
{
	char *qname;

	/* If we memset here, we lose our reference to the kmalloc'd
	 * res->lockname.name, so be sure to init every field
	 * correctly! */

	qname = (char *) res->lockname.name;
	memcpy(qname, name, namelen);

	res->lockname.len = namelen;
	res->lockname.hash = dlm_lockid_hash(name, namelen);

	init_waitqueue_head(&res->wq);
	spin_lock_init(&res->spinlock);
	INIT_HLIST_NODE(&res->hash_node);
	INIT_LIST_HEAD(&res->granted);
	INIT_LIST_HEAD(&res->converting);
	INIT_LIST_HEAD(&res->blocked);
	INIT_LIST_HEAD(&res->dirty);
	INIT_LIST_HEAD(&res->recovering);
	INIT_LIST_HEAD(&res->purge);
	atomic_set(&res->asts_reserved, 0);
	res->migration_pending = 0;

	kref_init(&res->refs);

	/* just for consistency */
	spin_lock(&res->spinlock);
	dlm_set_lockres_owner(dlm, res, DLM_LOCK_RES_OWNER_UNKNOWN);
	spin_unlock(&res->spinlock);

	res->state = DLM_LOCK_RES_IN_PROGRESS;

	res->last_used = 0;

	memset(res->lvb, 0, DLM_LVB_LEN);
}

struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm,
				   const char *name,
				   unsigned int namelen)
{
	struct dlm_lock_resource *res;

	res = kmalloc(sizeof(struct dlm_lock_resource), GFP_NOFS);
	if (!res)
		return NULL;

	res->lockname.name = kmalloc(namelen, GFP_NOFS);
	if (!res->lockname.name) {
		kfree(res);
		return NULL;
	}

	dlm_init_lockres(dlm, res, name, namelen);
	return res;
}

/*
 * lookup a lock resource by name.
 * may already exist in the hashtable.
 * lockid is null terminated
 *
 * if not, allocate enough for the lockres and for
 * the temporary structure used in doing the mastering.
 *
 * also, do a lookup in the dlm->master_list to see
 * if another node has begun mastering the same lock.
 * if so, there should be a block entry in there
 * for this name, and we should *not* attempt to master
 * the lock here.   need to wait around for that node
 * to assert_master (or die).
 *
 */
struct dlm_lock_resource * dlm_get_lock_resource(struct dlm_ctxt *dlm,
					  const char *lockid,
					  int namelen,
					  int flags)
{
	struct dlm_lock_resource *tmpres=NULL, *res=NULL;
	struct dlm_master_list_entry *mle = NULL;
	struct dlm_master_list_entry *alloc_mle = NULL;
	int blocked = 0;
	int ret, nodenum;
	struct dlm_node_iter iter;
	unsigned int hash;
	int tries = 0;
	int bit, wait_on_recovery = 0;

	BUG_ON(!lockid);

	hash = dlm_lockid_hash(lockid, namelen);

	mlog(0, "get lockres %s (len %d)\n", lockid, namelen);

lookup:
	spin_lock(&dlm->spinlock);
	tmpres = __dlm_lookup_lockres(dlm, lockid, namelen, hash);
	if (tmpres) {
		spin_unlock(&dlm->spinlock);
		mlog(0, "found in hash!\n");
		if (res)
			dlm_lockres_put(res);
		res = tmpres;
		goto leave;
	}

	if (!res) {
		spin_unlock(&dlm->spinlock);
		mlog(0, "allocating a new resource\n");
		/* nothing found and we need to allocate one. */
		alloc_mle = (struct dlm_master_list_entry *)
			kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
		if (!alloc_mle)
			goto leave;
		res = dlm_new_lockres(dlm, lockid, namelen);
		if (!res)
			goto leave;
		goto lookup;
	}

	mlog(0, "no lockres found, allocated our own: %p\n", res);

	if (flags & LKM_LOCAL) {
		/* caller knows it's safe to assume it's not mastered elsewhere
		 * DONE!  return right away */
		spin_lock(&res->spinlock);
		dlm_change_lockres_owner(dlm, res, dlm->node_num);
		__dlm_insert_lockres(dlm, res);
		spin_unlock(&res->spinlock);
		spin_unlock(&dlm->spinlock);
		/* lockres still marked IN_PROGRESS */
		goto wake_waiters;
	}

	/* check master list to see if another node has started mastering it */
	spin_lock(&dlm->master_lock);

	/* if we found a block, wait for lock to be mastered by another node */
	blocked = dlm_find_mle(dlm, &mle, (char *)lockid, namelen);
	if (blocked) {
		if (mle->type == DLM_MLE_MASTER) {
			mlog(ML_ERROR, "master entry for nonexistent lock!\n");
			BUG();
		} else if (mle->type == DLM_MLE_MIGRATION) {
			/* migration is in progress! */
			/* the good news is that we now know the
			 * "current" master (mle->master). */

			spin_unlock(&dlm->master_lock);
			assert_spin_locked(&dlm->spinlock);

			/* set the lockres owner and hash it */
			spin_lock(&res->spinlock);
			dlm_set_lockres_owner(dlm, res, mle->master);
			__dlm_insert_lockres(dlm, res);
			spin_unlock(&res->spinlock);
			spin_unlock(&dlm->spinlock);

			/* master is known, detach */
			dlm_mle_detach_hb_events(dlm, mle);
			dlm_put_mle(mle);
			mle = NULL;
			goto wake_waiters;
		}
	} else {
		/* go ahead and try to master lock on this node */
		mle = alloc_mle;
		/* make sure this does not get freed below */
		alloc_mle = NULL;
		dlm_init_mle(mle, DLM_MLE_MASTER, dlm, res, NULL, 0);
		set_bit(dlm->node_num, mle->maybe_map);
		list_add(&mle->list, &dlm->master_list);

		/* still holding the dlm spinlock, check the recovery map
		 * to see if there are any nodes that still need to be 
		 * considered.  these will not appear in the mle nodemap
		 * but they might own this lockres.  wait on them. */
		bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
		if (bit < O2NM_MAX_NODES) {
			mlog(ML_NOTICE, "%s:%.*s: at least one node (%d) to"
			     "recover before lock mastery can begin\n",
			     dlm->name, namelen, (char *)lockid, bit);
			wait_on_recovery = 1;
		}
	}

	/* at this point there is either a DLM_MLE_BLOCK or a
	 * DLM_MLE_MASTER on the master list, so it's safe to add the
	 * lockres to the hashtable.  anyone who finds the lock will
	 * still have to wait on the IN_PROGRESS. */

	/* finally add the lockres to its hash bucket */
	__dlm_insert_lockres(dlm, res);
	/* get an extra ref on the mle in case this is a BLOCK
	 * if so, the creator of the BLOCK may try to put the last
	 * ref at this time in the assert master handler, so we
	 * need an extra one to keep from a bad ptr deref. */
	dlm_get_mle_inuse(mle);
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);

redo_request:
	while (wait_on_recovery) {
		/* any cluster changes that occurred after dropping the
		 * dlm spinlock would be detectable be a change on the mle,
		 * so we only need to clear out the recovery map once. */
		if (dlm_is_recovery_lock(lockid, namelen)) {
			mlog(ML_NOTICE, "%s: recovery map is not empty, but "
			     "must master $RECOVERY lock now\n", dlm->name);
			if (!dlm_pre_master_reco_lockres(dlm, res))
				wait_on_recovery = 0;
			else {
				mlog(0, "%s: waiting 500ms for heartbeat state "
				    "change\n", dlm->name);
				msleep(500);
			}
			continue;
		} 

		dlm_kick_recovery_thread(dlm);
		msleep(1000);
		dlm_wait_for_recovery(dlm);

		spin_lock(&dlm->spinlock);
		bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
		if (bit < O2NM_MAX_NODES) {
			mlog(ML_NOTICE, "%s:%.*s: at least one node (%d) to"
			     "recover before lock mastery can begin\n",
			     dlm->name, namelen, (char *)lockid, bit);
			wait_on_recovery = 1;
		} else
			wait_on_recovery = 0;
		spin_unlock(&dlm->spinlock);

		if (wait_on_recovery)
			dlm_wait_for_node_recovery(dlm, bit, 10000);
	}

	/* must wait for lock to be mastered elsewhere */
	if (blocked)
		goto wait;

	ret = -EINVAL;
	dlm_node_iter_init(mle->vote_map, &iter);
	while ((nodenum = dlm_node_iter_next(&iter)) >= 0) {
		ret = dlm_do_master_request(mle, nodenum);
		if (ret < 0)
			mlog_errno(ret);
		if (mle->master != O2NM_MAX_NODES) {
			/* found a master ! */
			if (mle->master <= nodenum)
				break;
			/* if our master request has not reached the master
			 * yet, keep going until it does.  this is how the
			 * master will know that asserts are needed back to
			 * the lower nodes. */
			mlog(0, "%s:%.*s: requests only up to %u but master "
			     "is %u, keep going\n", dlm->name, namelen,
			     lockid, nodenum, mle->master);
		}
	}

wait:
	/* keep going until the response map includes all nodes */
	ret = dlm_wait_for_lock_mastery(dlm, res, mle, &blocked);
	if (ret < 0) {
		wait_on_recovery = 1;
		mlog(0, "%s:%.*s: node map changed, redo the "
		     "master request now, blocked=%d\n",
		     dlm->name, res->lockname.len,
		     res->lockname.name, blocked);
		if (++tries > 20) {
			mlog(ML_ERROR, "%s:%.*s: spinning on "
			     "dlm_wait_for_lock_mastery, blocked=%d\n", 
			     dlm->name, res->lockname.len, 
			     res->lockname.name, blocked);
			dlm_print_one_lock_resource(res);
			dlm_print_one_mle(mle);
			tries = 0;
		}
		goto redo_request;
	}

	mlog(0, "lockres mastered by %u\n", res->owner);
	/* make sure we never continue without this */
	BUG_ON(res->owner == O2NM_MAX_NODES);

	/* master is known, detach if not already detached */
	dlm_mle_detach_hb_events(dlm, mle);
	dlm_put_mle(mle);
	/* put the extra ref */
	dlm_put_mle_inuse(mle);

wake_waiters:
	spin_lock(&res->spinlock);
	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
	spin_unlock(&res->spinlock);
	wake_up(&res->wq);

leave:
	/* need to free the unused mle */
	if (alloc_mle)
		kmem_cache_free(dlm_mle_cache, alloc_mle);

	return res;
}


#define DLM_MASTERY_TIMEOUT_MS   5000

static int dlm_wait_for_lock_mastery(struct dlm_ctxt *dlm,
				     struct dlm_lock_resource *res,
				     struct dlm_master_list_entry *mle,
				     int *blocked)
{
	u8 m;
	int ret, bit;
	int map_changed, voting_done;
	int assert, sleep;

recheck:
	ret = 0;
	assert = 0;

	/* check if another node has already become the owner */
	spin_lock(&res->spinlock);
	if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
		mlog(0, "%s:%.*s: owner is suddenly %u\n", dlm->name,
		     res->lockname.len, res->lockname.name, res->owner);
		spin_unlock(&res->spinlock);
		/* this will cause the master to re-assert across
		 * the whole cluster, freeing up mles */
		if (res->owner != dlm->node_num) {
			ret = dlm_do_master_request(mle, res->owner);
			if (ret < 0) {
				/* give recovery a chance to run */
				mlog(ML_ERROR, "link to %u went down?: %d\n", res->owner, ret);
				msleep(500);
				goto recheck;
			}
		}
		ret = 0;
		goto leave;
	}
	spin_unlock(&res->spinlock);

	spin_lock(&mle->spinlock);
	m = mle->master;
	map_changed = (memcmp(mle->vote_map, mle->node_map,
			      sizeof(mle->vote_map)) != 0);
	voting_done = (memcmp(mle->vote_map, mle->response_map,
			     sizeof(mle->vote_map)) == 0);

	/* restart if we hit any errors */
	if (map_changed) {
		int b;
		mlog(0, "%s: %.*s: node map changed, restarting\n",
		     dlm->name, res->lockname.len, res->lockname.name);
		ret = dlm_restart_lock_mastery(dlm, res, mle, *blocked);
		b = (mle->type == DLM_MLE_BLOCK);
		if ((*blocked && !b) || (!*blocked && b)) {
			mlog(0, "%s:%.*s: status change: old=%d new=%d\n", 
			     dlm->name, res->lockname.len, res->lockname.name,
			     *blocked, b);
			*blocked = b;
		}
		spin_unlock(&mle->spinlock);
		if (ret < 0) {
			mlog_errno(ret);
			goto leave;
		}
		mlog(0, "%s:%.*s: restart lock mastery succeeded, "
		     "rechecking now\n", dlm->name, res->lockname.len,
		     res->lockname.name);
		goto recheck;
	} else {
		if (!voting_done) {
			mlog(0, "map not changed and voting not done "
			     "for %s:%.*s\n", dlm->name, res->lockname.len,
			     res->lockname.name);
		}
	}

	if (m != O2NM_MAX_NODES) {
		/* another node has done an assert!
		 * all done! */
		sleep = 0;
	} else {
		sleep = 1;
		/* have all nodes responded? */
		if (voting_done && !*blocked) {
			bit = find_next_bit(mle->maybe_map, O2NM_MAX_NODES, 0);
			if (dlm->node_num <= bit) {
				/* my node number is lowest.
			 	 * now tell other nodes that I am
				 * mastering this. */
				mle->master = dlm->node_num;
				assert = 1;
				sleep = 0;
			}
			/* if voting is done, but we have not received
			 * an assert master yet, we must sleep */
		}
	}

	spin_unlock(&mle->spinlock);

	/* sleep if we haven't finished voting yet */
	if (sleep) {
		unsigned long timeo = msecs_to_jiffies(DLM_MASTERY_TIMEOUT_MS);

		/*
		if (atomic_read(&mle->mle_refs.refcount) < 2)
			mlog(ML_ERROR, "mle (%p) refs=%d, name=%.*s\n", mle,
			atomic_read(&mle->mle_refs.refcount),
			res->lockname.len, res->lockname.name);
		*/
		atomic_set(&mle->woken, 0);
		(void)wait_event_timeout(mle->wq,
					 (atomic_read(&mle->woken) == 1),
					 timeo);
		if (res->owner == O2NM_MAX_NODES) {
			mlog(0, "waiting again\n");
			goto recheck;
		}
		mlog(0, "done waiting, master is %u\n", res->owner);
		ret = 0;
		goto leave;
	}

	ret = 0;   /* done */
	if (assert) {
		m = dlm->node_num;
		mlog(0, "about to master %.*s here, this=%u\n",
		     res->lockname.len, res->lockname.name, m);
		ret = dlm_do_assert_master(dlm, res->lockname.name,
					   res->lockname.len, mle->vote_map, 0);
		if (ret) {
			/* This is a failure in the network path,
			 * not in the response to the assert_master
			 * (any nonzero response is a BUG on this node).
			 * Most likely a socket just got disconnected
			 * due to node death. */
			mlog_errno(ret);
		}
		/* no longer need to restart lock mastery.
		 * all living nodes have been contacted. */
		ret = 0;
	}

	/* set the lockres owner */
	spin_lock(&res->spinlock);
	dlm_change_lockres_owner(dlm, res, m);
	spin_unlock(&res->spinlock);

leave:
	return ret;
}

struct dlm_bitmap_diff_iter
{
	int curnode;
	unsigned long *orig_bm;
	unsigned long *cur_bm;
	unsigned long diff_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
};

enum dlm_node_state_change
{
	NODE_DOWN = -1,
	NODE_NO_CHANGE = 0,
	NODE_UP
};

static void dlm_bitmap_diff_iter_init(struct dlm_bitmap_diff_iter *iter,
				      unsigned long *orig_bm,
				      unsigned long *cur_bm)
{
	unsigned long p1, p2;
	int i;

	iter->curnode = -1;
	iter->orig_bm = orig_bm;
	iter->cur_bm = cur_bm;

	for (i = 0; i < BITS_TO_LONGS(O2NM_MAX_NODES); i++) {
       		p1 = *(iter->orig_bm + i);
	       	p2 = *(iter->cur_bm + i);
		iter->diff_bm[i] = (p1 & ~p2) | (p2 & ~p1);
	}
}

static int dlm_bitmap_diff_iter_next(struct dlm_bitmap_diff_iter *iter,
				     enum dlm_node_state_change *state)
{
	int bit;

	if (iter->curnode >= O2NM_MAX_NODES)
		return -ENOENT;

	bit = find_next_bit(iter->diff_bm, O2NM_MAX_NODES,
			    iter->curnode+1);
	if (bit >= O2NM_MAX_NODES) {
		iter->curnode = O2NM_MAX_NODES;
		return -ENOENT;
	}

	/* if it was there in the original then this node died */
	if (test_bit(bit, iter->orig_bm))
		*state = NODE_DOWN;
	else
		*state = NODE_UP;

	iter->curnode = bit;
	return bit;
}


static int dlm_restart_lock_mastery(struct dlm_ctxt *dlm,
				    struct dlm_lock_resource *res,
				    struct dlm_master_list_entry *mle,
				    int blocked)
{
	struct dlm_bitmap_diff_iter bdi;
	enum dlm_node_state_change sc;
	int node;
	int ret = 0;

	mlog(0, "something happened such that the "
	     "master process may need to be restarted!\n");

	assert_spin_locked(&mle->spinlock);

	dlm_bitmap_diff_iter_init(&bdi, mle->vote_map, mle->node_map);
	node = dlm_bitmap_diff_iter_next(&bdi, &sc);
	while (node >= 0) {
		if (sc == NODE_UP) {
			/* a node came up.  clear any old vote from
			 * the response map and set it in the vote map
			 * then restart the mastery. */
			mlog(ML_NOTICE, "node %d up while restarting\n", node);

			/* redo the master request, but only for the new node */
			mlog(0, "sending request to new node\n");
			clear_bit(node, mle->response_map);
			set_bit(node, mle->vote_map);
		} else {
			mlog(ML_ERROR, "node down! %d\n", node);
			if (blocked) {
				int lowest = find_next_bit(mle->maybe_map,
						       O2NM_MAX_NODES, 0);

				/* act like it was never there */
				clear_bit(node, mle->maybe_map);

			       	if (node == lowest) {
					mlog(0, "expected master %u died"
					    " while this node was blocked "
					    "waiting on it!\n", node);
					lowest = find_next_bit(mle->maybe_map,
						       	O2NM_MAX_NODES,
						       	lowest+1);
					if (lowest < O2NM_MAX_NODES) {
						mlog(0, "%s:%.*s:still "
						     "blocked. waiting on %u "
						     "now\n", dlm->name,
						     res->lockname.len,
						     res->lockname.name,
						     lowest);
					} else {
						/* mle is an MLE_BLOCK, but
						 * there is now nothing left to
						 * block on.  we need to return
						 * all the way back out and try
						 * again with an MLE_MASTER.
						 * dlm_do_local_recovery_cleanup
						 * has already run, so the mle
						 * refcount is ok */
						mlog(0, "%s:%.*s: no "
						     "longer blocking. try to "
						     "master this here\n",
						     dlm->name,
						     res->lockname.len,
						     res->lockname.name);
						mle->type = DLM_MLE_MASTER;
						mle->u.res = res;
					}
				}
			}

			/* now blank out everything, as if we had never
			 * contacted anyone */
			memset(mle->maybe_map, 0, sizeof(mle->maybe_map));
			memset(mle->response_map, 0, sizeof(mle->response_map));
			/* reset the vote_map to the current node_map */
			memcpy(mle->vote_map, mle->node_map,
			       sizeof(mle->node_map));
			/* put myself into the maybe map */
			if (mle->type != DLM_MLE_BLOCK)
				set_bit(dlm->node_num, mle->maybe_map);
		}
		ret = -EAGAIN;
		node = dlm_bitmap_diff_iter_next(&bdi, &sc);
	}
	return ret;
}


/*
 * DLM_MASTER_REQUEST_MSG
 *
 * returns: 0 on success,
 *          -errno on a network error
 *
 * on error, the caller should assume the target node is "dead"
 *
 */

static int dlm_do_master_request(struct dlm_master_list_entry *mle, int to)
{
	struct dlm_ctxt *dlm = mle->dlm;
	struct dlm_master_request request;
	int ret, response=0, resend;

	memset(&request, 0, sizeof(request));
	request.node_idx = dlm->node_num;

	BUG_ON(mle->type == DLM_MLE_MIGRATION);

	if (mle->type != DLM_MLE_MASTER) {
		request.namelen = mle->u.name.len;
		memcpy(request.name, mle->u.name.name, request.namelen);
	} else {
		request.namelen = mle->u.res->lockname.len;
		memcpy(request.name, mle->u.res->lockname.name,
			request.namelen);
	}

again:
	ret = o2net_send_message(DLM_MASTER_REQUEST_MSG, dlm->key, &request,
				 sizeof(request), to, &response);
	if (ret < 0)  {
		if (ret == -ESRCH) {
			/* should never happen */
			mlog(ML_ERROR, "TCP stack not ready!\n");
			BUG();
		} else if (ret == -EINVAL) {
			mlog(ML_ERROR, "bad args passed to o2net!\n");
			BUG();
		} else if (ret == -ENOMEM) {
			mlog(ML_ERROR, "out of memory while trying to send "
			     "network message!  retrying\n");
			/* this is totally crude */
			msleep(50);
			goto again;
		} else if (!dlm_is_host_down(ret)) {
			/* not a network error. bad. */
			mlog_errno(ret);
			mlog(ML_ERROR, "unhandled error!");
			BUG();
		}
		/* all other errors should be network errors,
		 * and likely indicate node death */
		mlog(ML_ERROR, "link to %d went down!\n", to);
		goto out;
	}

	ret = 0;
	resend = 0;
	spin_lock(&mle->spinlock);
	switch (response) {
		case DLM_MASTER_RESP_YES:
			set_bit(to, mle->response_map);
			mlog(0, "node %u is the master, response=YES\n", to);
			mle->master = to;
			break;
		case DLM_MASTER_RESP_NO:
			mlog(0, "node %u not master, response=NO\n", to);
			set_bit(to, mle->response_map);
			break;
		case DLM_MASTER_RESP_MAYBE:
			mlog(0, "node %u not master, response=MAYBE\n", to);
			set_bit(to, mle->response_map);
			set_bit(to, mle->maybe_map);
			break;
		case DLM_MASTER_RESP_ERROR:
			mlog(0, "node %u hit an error, resending\n", to);
			resend = 1;
			response = 0;
			break;
		default:
			mlog(ML_ERROR, "bad response! %u\n", response);
			BUG();
	}
	spin_unlock(&mle->spinlock);
	if (resend) {
		/* this is also totally crude */
		msleep(50);
		goto again;
	}

out:
	return ret;
}

/*
 * locks that can be taken here:
 * dlm->spinlock
 * res->spinlock
 * mle->spinlock
 * dlm->master_list
 *
 * if possible, TRIM THIS DOWN!!!
 */
int dlm_master_request_handler(struct o2net_msg *msg, u32 len, void *data)
{
	u8 response = DLM_MASTER_RESP_MAYBE;
	struct dlm_ctxt *dlm = data;
	struct dlm_lock_resource *res = NULL;
	struct dlm_master_request *request = (struct dlm_master_request *) msg->buf;
	struct dlm_master_list_entry *mle = NULL, *tmpmle = NULL;
	char *name;
	unsigned int namelen, hash;
	int found, ret;
	int set_maybe;
	int dispatch_assert = 0;

	if (!dlm_grab(dlm))
		return DLM_MASTER_RESP_NO;

	if (!dlm_domain_fully_joined(dlm)) {
		response = DLM_MASTER_RESP_NO;
		goto send_response;
	}

	name = request->name;
	namelen = request->namelen;
	hash = dlm_lockid_hash(name, namelen);

	if (namelen > DLM_LOCKID_NAME_MAX) {
		response = DLM_IVBUFLEN;
		goto send_response;
	}

way_up_top:
	spin_lock(&dlm->spinlock);
	res = __dlm_lookup_lockres(dlm, name, namelen, hash);
	if (res) {
		spin_unlock(&dlm->spinlock);

		/* take care of the easy cases up front */
		spin_lock(&res->spinlock);
		if (res->state & DLM_LOCK_RES_RECOVERING) {
			spin_unlock(&res->spinlock);
			mlog(0, "returning DLM_MASTER_RESP_ERROR since res is "
			     "being recovered\n");
			response = DLM_MASTER_RESP_ERROR;
			if (mle)
				kmem_cache_free(dlm_mle_cache, mle);
			goto send_response;
		}

		if (res->owner == dlm->node_num) {
			spin_unlock(&res->spinlock);
			// mlog(0, "this node is the master\n");
			response = DLM_MASTER_RESP_YES;
			if (mle)
				kmem_cache_free(dlm_mle_cache, mle);

			/* this node is the owner.
			 * there is some extra work that needs to
			 * happen now.  the requesting node has
			 * caused all nodes up to this one to
			 * create mles.  this node now needs to
			 * go back and clean those up. */
			dispatch_assert = 1;
			goto send_response;
		} else if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
			spin_unlock(&res->spinlock);
			// mlog(0, "node %u is the master\n", res->owner);
			response = DLM_MASTER_RESP_NO;
			if (mle)
				kmem_cache_free(dlm_mle_cache, mle);
			goto send_response;
		}

		/* ok, there is no owner.  either this node is
		 * being blocked, or it is actively trying to
		 * master this lock. */
		if (!(res->state & DLM_LOCK_RES_IN_PROGRESS)) {
			mlog(ML_ERROR, "lock with no owner should be "
			     "in-progress!\n");
			BUG();
		}

		// mlog(0, "lockres is in progress...\n");
		spin_lock(&dlm->master_lock);
		found = dlm_find_mle(dlm, &tmpmle, name, namelen);
		if (!found) {
			mlog(ML_ERROR, "no mle found for this lock!\n");
			BUG();
		}
		set_maybe = 1;
		spin_lock(&tmpmle->spinlock);
		if (tmpmle->type == DLM_MLE_BLOCK) {
			// mlog(0, "this node is waiting for "
			// "lockres to be mastered\n");
			response = DLM_MASTER_RESP_NO;
		} else if (tmpmle->type == DLM_MLE_MIGRATION) {
			mlog(0, "node %u is master, but trying to migrate to "
			     "node %u.\n", tmpmle->master, tmpmle->new_master);
			if (tmpmle->master == dlm->node_num) {
				response = DLM_MASTER_RESP_YES;
				mlog(ML_ERROR, "no owner on lockres, but this "
				     "node is trying to migrate it to %u?!\n",
				     tmpmle->new_master);
				BUG();
			} else {
				/* the real master can respond on its own */
				response = DLM_MASTER_RESP_NO;
			}
		} else if (tmpmle->master != DLM_LOCK_RES_OWNER_UNKNOWN) {
			set_maybe = 0;
			if (tmpmle->master == dlm->node_num) {
				response = DLM_MASTER_RESP_YES;
				/* this node will be the owner.
				 * go back and clean the mles on any
				 * other nodes */
				dispatch_assert = 1;
			} else
				response = DLM_MASTER_RESP_NO;
		} else {
			// mlog(0, "this node is attempting to "
			// "master lockres\n");
			response = DLM_MASTER_RESP_MAYBE;
		}
		if (set_maybe)
			set_bit(request->node_idx, tmpmle->maybe_map);
		spin_unlock(&tmpmle->spinlock);

		spin_unlock(&dlm->master_lock);
		spin_unlock(&res->spinlock);

		/* keep the mle attached to heartbeat events */
		dlm_put_mle(tmpmle);
		if (mle)
			kmem_cache_free(dlm_mle_cache, mle);
		goto send_response;
	}

	/*
	 * lockres doesn't exist on this node
	 * if there is an MLE_BLOCK, return NO
	 * if there is an MLE_MASTER, return MAYBE
	 * otherwise, add an MLE_BLOCK, return NO
	 */
	spin_lock(&dlm->master_lock);
	found = dlm_find_mle(dlm, &tmpmle, name, namelen);
	if (!found) {
		/* this lockid has never been seen on this node yet */
		// mlog(0, "no mle found\n");
		if (!mle) {
			spin_unlock(&dlm->master_lock);
			spin_unlock(&dlm->spinlock);

			mle = (struct dlm_master_list_entry *)
				kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
			if (!mle) {
				response = DLM_MASTER_RESP_ERROR;
				mlog_errno(-ENOMEM);
				goto send_response;
			}
			goto way_up_top;
		}

		// mlog(0, "this is second time thru, already allocated, "
		// "add the block.\n");
		dlm_init_mle(mle, DLM_MLE_BLOCK, dlm, NULL, name, namelen);
		set_bit(request->node_idx, mle->maybe_map);
		list_add(&mle->list, &dlm->master_list);
		response = DLM_MASTER_RESP_NO;
	} else {
		// mlog(0, "mle was found\n");
		set_maybe = 1;
		spin_lock(&tmpmle->spinlock);
		if (tmpmle->master == dlm->node_num) {
			mlog(ML_ERROR, "no lockres, but an mle with this node as master!\n");
			BUG();
		}
		if (tmpmle->type == DLM_MLE_BLOCK)
			response = DLM_MASTER_RESP_NO;
		else if (tmpmle->type == DLM_MLE_MIGRATION) {
			mlog(0, "migration mle was found (%u->%u)\n",
			     tmpmle->master, tmpmle->new_master);
			/* real master can respond on its own */
			response = DLM_MASTER_RESP_NO;
		} else
			response = DLM_MASTER_RESP_MAYBE;
		if (set_maybe)
			set_bit(request->node_idx, tmpmle->maybe_map);
		spin_unlock(&tmpmle->spinlock);
	}
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);

	if (found) {
		/* keep the mle attached to heartbeat events */
		dlm_put_mle(tmpmle);
	}
send_response:

	if (dispatch_assert) {
		if (response != DLM_MASTER_RESP_YES)
			mlog(ML_ERROR, "invalid response %d\n", response);
		if (!res) {
			mlog(ML_ERROR, "bad lockres while trying to assert!\n");
			BUG();
		}
		mlog(0, "%u is the owner of %.*s, cleaning everyone else\n",
			     dlm->node_num, res->lockname.len, res->lockname.name);
		ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx, 
						 DLM_ASSERT_MASTER_MLE_CLEANUP);
		if (ret < 0) {
			mlog(ML_ERROR, "failed to dispatch assert master work\n");
			response = DLM_MASTER_RESP_ERROR;
		}
	}

	dlm_put(dlm);
	return response;
}

/*
 * DLM_ASSERT_MASTER_MSG
 */


/*
 * NOTE: this can be used for debugging
 * can periodically run all locks owned by this node
 * and re-assert across the cluster...
 */
static int dlm_do_assert_master(struct dlm_ctxt *dlm, const char *lockname,
				unsigned int namelen, void *nodemap,
				u32 flags)
{
	struct dlm_assert_master assert;
	int to, tmpret;
	struct dlm_node_iter iter;
	int ret = 0;
	int reassert;

	BUG_ON(namelen > O2NM_MAX_NAME_LEN);
again:
	reassert = 0;

	/* note that if this nodemap is empty, it returns 0 */
	dlm_node_iter_init(nodemap, &iter);
	while ((to = dlm_node_iter_next(&iter)) >= 0) {
		int r = 0;
		struct dlm_master_list_entry *mle = NULL;

		mlog(0, "sending assert master to %d (%.*s)\n", to,
		     namelen, lockname);
		memset(&assert, 0, sizeof(assert));
		assert.node_idx = dlm->node_num;
		assert.namelen = namelen;
		memcpy(assert.name, lockname, namelen);
		assert.flags = cpu_to_be32(flags);

		tmpret = o2net_send_message(DLM_ASSERT_MASTER_MSG, dlm->key,
					    &assert, sizeof(assert), to, &r);
		if (tmpret < 0) {
			mlog(0, "assert_master returned %d!\n", tmpret);
			if (!dlm_is_host_down(tmpret)) {
				mlog(ML_ERROR, "unhandled error=%d!\n", tmpret);
				BUG();
			}
			/* a node died.  finish out the rest of the nodes. */
			mlog(0, "link to %d went down!\n", to);
			/* any nonzero status return will do */
			ret = tmpret;
		} else if (r < 0) {
			/* ok, something horribly messed.  kill thyself. */
			mlog(ML_ERROR,"during assert master of %.*s to %u, "
			     "got %d.\n", namelen, lockname, to, r);
			spin_lock(&dlm->spinlock);
			spin_lock(&dlm->master_lock);
			if (dlm_find_mle(dlm, &mle, (char *)lockname,
					 namelen)) {
				dlm_print_one_mle(mle);
				__dlm_put_mle(mle);
			}
			spin_unlock(&dlm->master_lock);
			spin_unlock(&dlm->spinlock);
			BUG();
		} else if (r == EAGAIN) {
			mlog(0, "%.*s: node %u create mles on other "
			     "nodes and requests a re-assert\n", 
			     namelen, lockname, to);
			reassert = 1;
		}
	}

	if (reassert)
		goto again;

	return ret;
}

/*
 * locks that can be taken here:
 * dlm->spinlock
 * res->spinlock
 * mle->spinlock
 * dlm->master_list
 *
 * if possible, TRIM THIS DOWN!!!
 */
int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data)
{
	struct dlm_ctxt *dlm = data;
	struct dlm_master_list_entry *mle = NULL;
	struct dlm_assert_master *assert = (struct dlm_assert_master *)msg->buf;
	struct dlm_lock_resource *res = NULL;
	char *name;
	unsigned int namelen, hash;
	u32 flags;
	int master_request = 0;
	int ret = 0;

	if (!dlm_grab(dlm))
		return 0;

	name = assert->name;
	namelen = assert->namelen;
	hash = dlm_lockid_hash(name, namelen);
	flags = be32_to_cpu(assert->flags);

	if (namelen > DLM_LOCKID_NAME_MAX) {
		mlog(ML_ERROR, "Invalid name length!");
		goto done;
	}

	spin_lock(&dlm->spinlock);

	if (flags)
		mlog(0, "assert_master with flags: %u\n", flags);

	/* find the MLE */
	spin_lock(&dlm->master_lock);
	if (!dlm_find_mle(dlm, &mle, name, namelen)) {
		/* not an error, could be master just re-asserting */
		mlog(0, "just got an assert_master from %u, but no "
		     "MLE for it! (%.*s)\n", assert->node_idx,
		     namelen, name);
	} else {
		int bit = find_next_bit (mle->maybe_map, O2NM_MAX_NODES, 0);
		if (bit >= O2NM_MAX_NODES) {
			/* not necessarily an error, though less likely.
			 * could be master just re-asserting. */
			mlog(0, "no bits set in the maybe_map, but %u "
			     "is asserting! (%.*s)\n", assert->node_idx,
			     namelen, name);
		} else if (bit != assert->node_idx) {
			if (flags & DLM_ASSERT_MASTER_MLE_CLEANUP) {
				mlog(0, "master %u was found, %u should "
				     "back off\n", assert->node_idx, bit);
			} else {
				/* with the fix for bug 569, a higher node
				 * number winning the mastery will respond
				 * YES to mastery requests, but this node
				 * had no way of knowing.  let it pass. */
				mlog(0, "%u is the lowest node, "
				     "%u is asserting. (%.*s)  %u must "
				     "have begun after %u won.\n", bit,
				     assert->node_idx, namelen, name, bit,
				     assert->node_idx);
			}
		}
		if (mle->type == DLM_MLE_MIGRATION) {
			if (flags & DLM_ASSERT_MASTER_MLE_CLEANUP) {
				mlog(0, "%s:%.*s: got cleanup assert"
				     " from %u for migration\n",
				     dlm->name, namelen, name,
				     assert->node_idx);
			} else if (!(flags & DLM_ASSERT_MASTER_FINISH_MIGRATION)) {
				mlog(0, "%s:%.*s: got unrelated assert"
				     " from %u for migration, ignoring\n",
				     dlm->name, namelen, name,
				     assert->node_idx);
				__dlm_put_mle(mle);
				spin_unlock(&dlm->master_lock);
				spin_unlock(&dlm->spinlock);
				goto done;
			}	
		}
	}
	spin_unlock(&dlm->master_lock);

	/* ok everything checks out with the MLE
	 * now check to see if there is a lockres */
	res = __dlm_lookup_lockres(dlm, name, namelen, hash);
	if (res) {
		spin_lock(&res->spinlock);
		if (res->state & DLM_LOCK_RES_RECOVERING)  {
			mlog(ML_ERROR, "%u asserting but %.*s is "
			     "RECOVERING!\n", assert->node_idx, namelen, name);
			goto kill;
		}
		if (!mle) {
			if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN &&
			    res->owner != assert->node_idx) {
				mlog(ML_ERROR, "assert_master from "
					  "%u, but current owner is "
					  "%u! (%.*s)\n",
				       assert->node_idx, res->owner,
				       namelen, name);
				goto kill;
			}
		} else if (mle->type != DLM_MLE_MIGRATION) {
			if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
				/* owner is just re-asserting */
				if (res->owner == assert->node_idx) {
					mlog(0, "owner %u re-asserting on "
					     "lock %.*s\n", assert->node_idx,
					     namelen, name);
					goto ok;
				}
				mlog(ML_ERROR, "got assert_master from "
				     "node %u, but %u is the owner! "
				     "(%.*s)\n", assert->node_idx,
				     res->owner, namelen, name);
				goto kill;
			}
			if (!(res->state & DLM_LOCK_RES_IN_PROGRESS)) {
				mlog(ML_ERROR, "got assert from %u, but lock "
				     "with no owner should be "
				     "in-progress! (%.*s)\n",
				     assert->node_idx,
				     namelen, name);
				goto kill;
			}
		} else /* mle->type == DLM_MLE_MIGRATION */ {
			/* should only be getting an assert from new master */
			if (assert->node_idx != mle->new_master) {
				mlog(ML_ERROR, "got assert from %u, but "
				     "new master is %u, and old master "
				     "was %u (%.*s)\n",
				     assert->node_idx, mle->new_master,
				     mle->master, namelen, name);
				goto kill;
			}

		}
ok:
		spin_unlock(&res->spinlock);
	}
	spin_unlock(&dlm->spinlock);

	// mlog(0, "woo!  got an assert_master from node %u!\n",
	// 	     assert->node_idx);
	if (mle) {
		int extra_ref = 0;
		int nn = -1;
		int rr, err = 0;
		
		spin_lock(&mle->spinlock);
		if (mle->type == DLM_MLE_BLOCK || mle->type == DLM_MLE_MIGRATION)
			extra_ref = 1;
		else {
			/* MASTER mle: if any bits set in the response map
			 * then the calling node needs to re-assert to clear
			 * up nodes that this node contacted */
			while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES, 
						    nn+1)) < O2NM_MAX_NODES) {
				if (nn != dlm->node_num && nn != assert->node_idx)
					master_request = 1;
			}
		}
		mle->master = assert->node_idx;
		atomic_set(&mle->woken, 1);
		wake_up(&mle->wq);
		spin_unlock(&mle->spinlock);

		if (res) {
			spin_lock(&res->spinlock);
			if (mle->type == DLM_MLE_MIGRATION) {
				mlog(0, "finishing off migration of lockres %.*s, "
			     		"from %u to %u\n",
			       		res->lockname.len, res->lockname.name,
			       		dlm->node_num, mle->new_master);
				res->state &= ~DLM_LOCK_RES_MIGRATING;
				dlm_change_lockres_owner(dlm, res, mle->new_master);
				BUG_ON(res->state & DLM_LOCK_RES_DIRTY);
			} else {
				dlm_change_lockres_owner(dlm, res, mle->master);
			}
			spin_unlock(&res->spinlock);
		}

		/* master is known, detach if not already detached.
		 * ensures that only one assert_master call will happen
		 * on this mle. */
		spin_lock(&dlm->spinlock);
		spin_lock(&dlm->master_lock);

		rr = atomic_read(&mle->mle_refs.refcount);
		if (mle->inuse > 0) {
			if (extra_ref && rr < 3)
				err = 1;
			else if (!extra_ref && rr < 2)
				err = 1;
		} else {
			if (extra_ref && rr < 2)
				err = 1;
			else if (!extra_ref && rr < 1)
				err = 1;
		}
		if (err) {
			mlog(ML_ERROR, "%s:%.*s: got assert master from %u "
			     "that will mess up this node, refs=%d, extra=%d, "
			     "inuse=%d\n", dlm->name, namelen, name,
			     assert->node_idx, rr, extra_ref, mle->inuse);
			dlm_print_one_mle(mle);
		}
		list_del_init(&mle->list);
		__dlm_mle_detach_hb_events(dlm, mle);
		__dlm_put_mle(mle);
		if (extra_ref) {
			/* the assert master message now balances the extra
		 	 * ref given by the master / migration request message.
		 	 * if this is the last put, it will be removed
		 	 * from the list. */
			__dlm_put_mle(mle);
		}
		spin_unlock(&dlm->master_lock);
		spin_unlock(&dlm->spinlock);
	} else if (res) {
		if (res->owner != assert->node_idx) {
			mlog(0, "assert_master from %u, but current "
			     "owner is %u (%.*s), no mle\n", assert->node_idx,
			     res->owner, namelen, name);
		}
	}

done:
	ret = 0;
	if (res)
		dlm_lockres_put(res);
	dlm_put(dlm);
	if (master_request) {
		mlog(0, "need to tell master to reassert\n");
		ret = EAGAIN;  // positive. negative would shoot down the node.
	}
	return ret;

kill:
	/* kill the caller! */
	mlog(ML_ERROR, "Bad message received from another node.  Dumping state "
	     "and killing the other node now!  This node is OK and can continue.\n");
	__dlm_print_one_lock_resource(res);
	spin_unlock(&res->spinlock);
	spin_unlock(&dlm->spinlock);
	dlm_lockres_put(res);
	dlm_put(dlm);
	return -EINVAL;
}

int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
			       struct dlm_lock_resource *res,
			       int ignore_higher, u8 request_from, u32 flags)
{
	struct dlm_work_item *item;
	item = kcalloc(1, sizeof(*item), GFP_NOFS);
	if (!item)
		return -ENOMEM;


	/* queue up work for dlm_assert_master_worker */
	dlm_grab(dlm);  /* get an extra ref for the work item */
	dlm_init_work_item(dlm, item, dlm_assert_master_worker, NULL);
	item->u.am.lockres = res; /* already have a ref */
	/* can optionally ignore node numbers higher than this node */
	item->u.am.ignore_higher = ignore_higher;
	item->u.am.request_from = request_from;
	item->u.am.flags = flags;

	if (ignore_higher) 
		mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len, 
		     res->lockname.name);
		
	spin_lock(&dlm->work_lock);
	list_add_tail(&item->list, &dlm->work_list);
	spin_unlock(&dlm->work_lock);

	queue_work(dlm->dlm_worker, &dlm->dispatched_work);
	return 0;
}

static void dlm_assert_master_worker(struct dlm_work_item *item, void *data)
{
	struct dlm_ctxt *dlm = data;
	int ret = 0;
	struct dlm_lock_resource *res;
	unsigned long nodemap[BITS_TO_LONGS(O2NM_MAX_NODES)];
	int ignore_higher;
	int bit;
	u8 request_from;
	u32 flags;

	dlm = item->dlm;
	res = item->u.am.lockres;
	ignore_higher = item->u.am.ignore_higher;
	request_from = item->u.am.request_from;
	flags = item->u.am.flags;

	spin_lock(&dlm->spinlock);
	memcpy(nodemap, dlm->domain_map, sizeof(nodemap));
	spin_unlock(&dlm->spinlock);

	clear_bit(dlm->node_num, nodemap);
	if (ignore_higher) {
		/* if is this just to clear up mles for nodes below
		 * this node, do not send the message to the original
		 * caller or any node number higher than this */
		clear_bit(request_from, nodemap);
		bit = dlm->node_num;
		while (1) {
			bit = find_next_bit(nodemap, O2NM_MAX_NODES,
					    bit+1);
		       	if (bit >= O2NM_MAX_NODES)
				break;
			clear_bit(bit, nodemap);
		}
	}

	/*
	 * If we're migrating this lock to someone else, we are no
	 * longer allowed to assert out own mastery.  OTOH, we need to
	 * prevent migration from starting while we're still asserting
	 * our dominance.  The reserved ast delays migration.
	 */
	spin_lock(&res->spinlock);
	if (res->state & DLM_LOCK_RES_MIGRATING) {
		mlog(0, "Someone asked us to assert mastery, but we're "
		     "in the middle of migration.  Skipping assert, "
		     "the new master will handle that.\n");
		spin_unlock(&res->spinlock);
		goto put;
	} else
		__dlm_lockres_reserve_ast(res);
	spin_unlock(&res->spinlock);

	/* this call now finishes out the nodemap
	 * even if one or more nodes die */
	mlog(0, "worker about to master %.*s here, this=%u\n",
		     res->lockname.len, res->lockname.name, dlm->node_num);
	ret = dlm_do_assert_master(dlm, res->lockname.name,
				   res->lockname.len,
				   nodemap, flags);
	if (ret < 0) {
		/* no need to restart, we are done */
		if (!dlm_is_host_down(ret))
			mlog_errno(ret);
	}

	/* Ok, we've asserted ourselves.  Let's let migration start. */
	dlm_lockres_release_ast(dlm, res);

put:
	dlm_lockres_put(res);

	mlog(0, "finished with dlm_assert_master_worker\n");
}

/* SPECIAL CASE for the $RECOVERY lock used by the recovery thread.
 * We cannot wait for node recovery to complete to begin mastering this
 * lockres because this lockres is used to kick off recovery! ;-)
 * So, do a pre-check on all living nodes to see if any of those nodes
 * think that $RECOVERY is currently mastered by a dead node.  If so,
 * we wait a short time to allow that node to get notified by its own
 * heartbeat stack, then check again.  All $RECOVERY lock resources
 * mastered by dead nodes are purged when the hearbeat callback is 
 * fired, so we can know for sure that it is safe to continue once
 * the node returns a live node or no node.  */
static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
				       struct dlm_lock_resource *res)
{
	struct dlm_node_iter iter;
	int nodenum;
	int ret = 0;
	u8 master = DLM_LOCK_RES_OWNER_UNKNOWN;

	spin_lock(&dlm->spinlock);
	dlm_node_iter_init(dlm->domain_map, &iter);
	spin_unlock(&dlm->spinlock);

	while ((nodenum = dlm_node_iter_next(&iter)) >= 0) {
		/* do not send to self */
		if (nodenum == dlm->node_num)
			continue;
		ret = dlm_do_master_requery(dlm, res, nodenum, &master);
		if (ret < 0) {
			mlog_errno(ret);
			if (!dlm_is_host_down(ret))
				BUG();
			/* host is down, so answer for that node would be
			 * DLM_LOCK_RES_OWNER_UNKNOWN.  continue. */
			ret = 0;
		}

		if (master != DLM_LOCK_RES_OWNER_UNKNOWN) {
			/* check to see if this master is in the recovery map */
			spin_lock(&dlm->spinlock);
			if (test_bit(master, dlm->recovery_map)) {
				mlog(ML_NOTICE, "%s: node %u has not seen "
				     "node %u go down yet, and thinks the "
				     "dead node is mastering the recovery "
				     "lock.  must wait.\n", dlm->name,
				     nodenum, master);
				ret = -EAGAIN;
			}
			spin_unlock(&dlm->spinlock);
			mlog(0, "%s: reco lock master is %u\n", dlm->name, 
			     master);
			break;
		}
	}
	return ret;
}


/*
 * DLM_MIGRATE_LOCKRES
 */


int dlm_migrate_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
			u8 target)
{
	struct dlm_master_list_entry *mle = NULL;
	struct dlm_master_list_entry *oldmle = NULL;
 	struct dlm_migratable_lockres *mres = NULL;
	int ret = -EINVAL;
	const char *name;
	unsigned int namelen;
	int mle_added = 0;
	struct list_head *queue, *iter;
	int i;
	struct dlm_lock *lock;
	int empty = 1;

	if (!dlm_grab(dlm))
		return -EINVAL;

	name = res->lockname.name;
	namelen = res->lockname.len;

	mlog(0, "migrating %.*s to %u\n", namelen, name, target);

	/*
	 * ensure this lockres is a proper candidate for migration
	 */
	spin_lock(&res->spinlock);
	if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) {
		mlog(0, "cannot migrate lockres with unknown owner!\n");
		spin_unlock(&res->spinlock);
		goto leave;
	}
	if (res->owner != dlm->node_num) {
		mlog(0, "cannot migrate lockres this node doesn't own!\n");
		spin_unlock(&res->spinlock);
		goto leave;
	}
	mlog(0, "checking queues...\n");
	queue = &res->granted;
	for (i=0; i<3; i++) {
		list_for_each(iter, queue) {
			lock = list_entry (iter, struct dlm_lock, list);
			empty = 0;
			if (lock->ml.node == dlm->node_num) {
				mlog(0, "found a lock owned by this node "
				     "still on the %s queue!  will not "
				     "migrate this lockres\n",
				     i==0 ? "granted" :
				     (i==1 ? "converting" : "blocked"));
				spin_unlock(&res->spinlock);
				ret = -ENOTEMPTY;
				goto leave;
			}
		}
		queue++;
	}
	mlog(0, "all locks on this lockres are nonlocal.  continuing\n");
	spin_unlock(&res->spinlock);

	/* no work to do */
	if (empty) {
		mlog(0, "no locks were found on this lockres! done!\n");
		ret = 0;
		goto leave;
	}

	/*
	 * preallocate up front
	 * if this fails, abort
	 */

	ret = -ENOMEM;
	mres = (struct dlm_migratable_lockres *) __get_free_page(GFP_NOFS);
	if (!mres) {
		mlog_errno(ret);
		goto leave;
	}

	mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache,
								GFP_NOFS);
	if (!mle) {
		mlog_errno(ret);
		goto leave;
	}
	ret = 0;

	/*
	 * find a node to migrate the lockres to
	 */

	mlog(0, "picking a migration node\n");
	spin_lock(&dlm->spinlock);
	/* pick a new node */
	if (!test_bit(target, dlm->domain_map) ||
	    target >= O2NM_MAX_NODES) {
		target = dlm_pick_migration_target(dlm, res);
	}
	mlog(0, "node %u chosen for migration\n", target);

	if (target >= O2NM_MAX_NODES ||
	    !test_bit(target, dlm->domain_map)) {
		/* target chosen is not alive */
		ret = -EINVAL;
	}

	if (ret) {
		spin_unlock(&dlm->spinlock);
		goto fail;
	}

	mlog(0, "continuing with target = %u\n", target);

	/*
	 * clear any existing master requests and
	 * add the migration mle to the list
	 */
	spin_lock(&dlm->master_lock);
	ret = dlm_add_migration_mle(dlm, res, mle, &oldmle, name,
				    namelen, target, dlm->node_num);
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);

	if (ret == -EEXIST) {
		mlog(0, "another process is already migrating it\n");
		goto fail;
	}
	mle_added = 1;

	/*
	 * set the MIGRATING flag and flush asts
	 * if we fail after this we need to re-dirty the lockres
	 */
	if (dlm_mark_lockres_migrating(dlm, res, target) < 0) {
		mlog(ML_ERROR, "tried to migrate %.*s to %u, but "
		     "the target went down.\n", res->lockname.len,
		     res->lockname.name, target);
		spin_lock(&res->spinlock);
		res->state &= ~DLM_LOCK_RES_MIGRATING;
		spin_unlock(&res->spinlock);
		ret = -EINVAL;
	}

fail:
	if (oldmle) {
		/* master is known, detach if not already detached */
		dlm_mle_detach_hb_events(dlm, oldmle);
		dlm_put_mle(oldmle);
	}

	if (ret < 0) {
		if (mle_added) {
			dlm_mle_detach_hb_events(dlm, mle);
			dlm_put_mle(mle);
		} else if (mle) {
			kmem_cache_free(dlm_mle_cache, mle);
		}
		goto leave;
	}

	/*
	 * at this point, we have a migration target, an mle
	 * in the master list, and the MIGRATING flag set on
	 * the lockres
	 */


	/* get an extra reference on the mle.
	 * otherwise the assert_master from the new
	 * master will destroy this.
	 * also, make sure that all callers of dlm_get_mle
	 * take both dlm->spinlock and dlm->master_lock */
	spin_lock(&dlm->spinlock);
	spin_lock(&dlm->master_lock);
	dlm_get_mle_inuse(mle);
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);

	/* notify new node and send all lock state */
	/* call send_one_lockres with migration flag.
	 * this serves as notice to the target node that a
	 * migration is starting. */
	ret = dlm_send_one_lockres(dlm, res, mres, target,
				   DLM_MRES_MIGRATION);

	if (ret < 0) {
		mlog(0, "migration to node %u failed with %d\n",
		     target, ret);
		/* migration failed, detach and clean up mle */
		dlm_mle_detach_hb_events(dlm, mle);
		dlm_put_mle(mle);
		dlm_put_mle_inuse(mle);
		spin_lock(&res->spinlock);
		res->state &= ~DLM_LOCK_RES_MIGRATING;
		spin_unlock(&res->spinlock);
		goto leave;
	}

	/* at this point, the target sends a message to all nodes,
	 * (using dlm_do_migrate_request).  this node is skipped since
	 * we had to put an mle in the list to begin the process.  this
	 * node now waits for target to do an assert master.  this node
	 * will be the last one notified, ensuring that the migration
	 * is complete everywhere.  if the target dies while this is
	 * going on, some nodes could potentially see the target as the
	 * master, so it is important that my recovery finds the migration
	 * mle and sets the master to UNKNONWN. */


	/* wait for new node to assert master */
	while (1) {
		ret = wait_event_interruptible_timeout(mle->wq,
					(atomic_read(&mle->woken) == 1),
					msecs_to_jiffies(5000));

		if (ret >= 0) {
		       	if (atomic_read(&mle->woken) == 1 ||
			    res->owner == target)
				break;

			mlog(0, "timed out during migration\n");
			/* avoid hang during shutdown when migrating lockres 
			 * to a node which also goes down */
			if (dlm_is_node_dead(dlm, target)) {
				mlog(0, "%s:%.*s: expected migration "
				     "target %u is no longer up, restarting\n",
				     dlm->name, res->lockname.len,
				     res->lockname.name, target);
				ret = -ERESTARTSYS;
			}
		}
		if (ret == -ERESTARTSYS) {
			/* migration failed, detach and clean up mle */
			dlm_mle_detach_hb_events(dlm, mle);
			dlm_put_mle(mle);
			dlm_put_mle_inuse(mle);
			spin_lock(&res->spinlock);
			res->state &= ~DLM_LOCK_RES_MIGRATING;
			spin_unlock(&res->spinlock);
			goto leave;
		}
		/* TODO: if node died: stop, clean up, return error */
	}

	/* all done, set the owner, clear the flag */
	spin_lock(&res->spinlock);
	dlm_set_lockres_owner(dlm, res, target);
	res->state &= ~DLM_LOCK_RES_MIGRATING;
	dlm_remove_nonlocal_locks(dlm, res);
	spin_unlock(&res->spinlock);
	wake_up(&res->wq);

	/* master is known, detach if not already detached */
	dlm_mle_detach_hb_events(dlm, mle);
	dlm_put_mle_inuse(mle);
	ret = 0;

	dlm_lockres_calc_usage(dlm, res);

leave:
	/* re-dirty the lockres if we failed */
	if (ret < 0)
		dlm_kick_thread(dlm, res);

	/* TODO: cleanup */
	if (mres)
		free_page((unsigned long)mres);

	dlm_put(dlm);

	mlog(0, "returning %d\n", ret);
	return ret;
}

int dlm_lock_basts_flushed(struct dlm_ctxt *dlm, struct dlm_lock *lock)
{
	int ret;
	spin_lock(&dlm->ast_lock);
	spin_lock(&lock->spinlock);
	ret = (list_empty(&lock->bast_list) && !lock->bast_pending);
	spin_unlock(&lock->spinlock);
	spin_unlock(&dlm->ast_lock);
	return ret;
}

static int dlm_migration_can_proceed(struct dlm_ctxt *dlm,
				     struct dlm_lock_resource *res,
				     u8 mig_target)
{
	int can_proceed;
	spin_lock(&res->spinlock);
	can_proceed = !!(res->state & DLM_LOCK_RES_MIGRATING);
	spin_unlock(&res->spinlock);

	/* target has died, so make the caller break out of the 
	 * wait_event, but caller must recheck the domain_map */
	spin_lock(&dlm->spinlock);
	if (!test_bit(mig_target, dlm->domain_map))
		can_proceed = 1;
	spin_unlock(&dlm->spinlock);
	return can_proceed;
}

int dlm_lockres_is_dirty(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
{
	int ret;
	spin_lock(&res->spinlock);
	ret = !!(res->state & DLM_LOCK_RES_DIRTY);
	spin_unlock(&res->spinlock);
	return ret;
}


static int dlm_mark_lockres_migrating(struct dlm_ctxt *dlm,
				       struct dlm_lock_resource *res,
				       u8 target)
{
	int ret = 0;

	mlog(0, "dlm_mark_lockres_migrating: %.*s, from %u to %u\n",
	       res->lockname.len, res->lockname.name, dlm->node_num,
	       target);
	/* need to set MIGRATING flag on lockres.  this is done by
	 * ensuring that all asts have been flushed for this lockres. */
	spin_lock(&res->spinlock);
	BUG_ON(res->migration_pending);
	res->migration_pending = 1;
	/* strategy is to reserve an extra ast then release
	 * it below, letting the release do all of the work */
	__dlm_lockres_reserve_ast(res);
	spin_unlock(&res->spinlock);

	/* now flush all the pending asts.. hang out for a bit */
	dlm_kick_thread(dlm, res);
	wait_event(dlm->ast_wq, !dlm_lockres_is_dirty(dlm, res));
	dlm_lockres_release_ast(dlm, res);

	mlog(0, "about to wait on migration_wq, dirty=%s\n",
	       res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no");
	/* if the extra ref we just put was the final one, this
	 * will pass thru immediately.  otherwise, we need to wait
	 * for the last ast to finish. */
again:
	ret = wait_event_interruptible_timeout(dlm->migration_wq,
		   dlm_migration_can_proceed(dlm, res, target),
		   msecs_to_jiffies(1000));
	if (ret < 0) {
		mlog(0, "woken again: migrating? %s, dead? %s\n",
		       res->state & DLM_LOCK_RES_MIGRATING ? "yes":"no",
		       test_bit(target, dlm->domain_map) ? "no":"yes");
	} else {
		mlog(0, "all is well: migrating? %s, dead? %s\n",
		       res->state & DLM_LOCK_RES_MIGRATING ? "yes":"no",
		       test_bit(target, dlm->domain_map) ? "no":"yes");
	}
	if (!dlm_migration_can_proceed(dlm, res, target)) {
		mlog(0, "trying again...\n");
		goto again;
	}

	/* did the target go down or die? */
	spin_lock(&dlm->spinlock);
	if (!test_bit(target, dlm->domain_map)) {
		mlog(ML_ERROR, "aha. migration target %u just went down\n",
		     target);
		ret = -EHOSTDOWN;
	}
	spin_unlock(&dlm->spinlock);

	/*
	 * at this point:
	 *
	 *   o the DLM_LOCK_RES_MIGRATING flag is set
	 *   o there are no pending asts on this lockres
	 *   o all processes trying to reserve an ast on this
	 *     lockres must wait for the MIGRATING flag to clear
	 */
	return ret;
}

/* last step in the migration process.
 * original master calls this to free all of the dlm_lock
 * structures that used to be for other nodes. */
static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
				      struct dlm_lock_resource *res)
{
	struct list_head *iter, *iter2;
	struct list_head *queue = &res->granted;
	int i;
	struct dlm_lock *lock;

	assert_spin_locked(&res->spinlock);

	BUG_ON(res->owner == dlm->node_num);

	for (i=0; i<3; i++) {
		list_for_each_safe(iter, iter2, queue) {
			lock = list_entry (iter, struct dlm_lock, list);
			if (lock->ml.node != dlm->node_num) {
				mlog(0, "putting lock for node %u\n",
				     lock->ml.node);
				/* be extra careful */
				BUG_ON(!list_empty(&lock->ast_list));
				BUG_ON(!list_empty(&lock->bast_list));
				BUG_ON(lock->ast_pending);
				BUG_ON(lock->bast_pending);
				list_del_init(&lock->list);
				dlm_lock_put(lock);
			}
		}
		queue++;
	}
}

/* for now this is not too intelligent.  we will
 * need stats to make this do the right thing.
 * this just finds the first lock on one of the
 * queues and uses that node as the target. */
static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
				    struct dlm_lock_resource *res)
{
	int i;
	struct list_head *queue = &res->granted;
	struct list_head *iter;
	struct dlm_lock *lock;
	int nodenum;

	assert_spin_locked(&dlm->spinlock);

	spin_lock(&res->spinlock);
	for (i=0; i<3; i++) {
		list_for_each(iter, queue) {
			/* up to the caller to make sure this node
			 * is alive */
			lock = list_entry (iter, struct dlm_lock, list);
			if (lock->ml.node != dlm->node_num) {
				spin_unlock(&res->spinlock);
				return lock->ml.node;
			}
		}
		queue++;
	}
	spin_unlock(&res->spinlock);
	mlog(0, "have not found a suitable target yet! checking domain map\n");

	/* ok now we're getting desperate.  pick anyone alive. */
	nodenum = -1;
	while (1) {
		nodenum = find_next_bit(dlm->domain_map,
					O2NM_MAX_NODES, nodenum+1);
		mlog(0, "found %d in domain map\n", nodenum);
		if (nodenum >= O2NM_MAX_NODES)
			break;
		if (nodenum != dlm->node_num) {
			mlog(0, "picking %d\n", nodenum);
			return nodenum;
		}
	}

	mlog(0, "giving up.  no master to migrate to\n");
	return DLM_LOCK_RES_OWNER_UNKNOWN;
}



/* this is called by the new master once all lockres
 * data has been received */
static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
				  struct dlm_lock_resource *res,
				  u8 master, u8 new_master,
				  struct dlm_node_iter *iter)
{
	struct dlm_migrate_request migrate;
	int ret, status = 0;
	int nodenum;

	memset(&migrate, 0, sizeof(migrate));
	migrate.namelen = res->lockname.len;
	memcpy(migrate.name, res->lockname.name, migrate.namelen);
	migrate.new_master = new_master;
	migrate.master = master;

	ret = 0;

	/* send message to all nodes, except the master and myself */
	while ((nodenum = dlm_node_iter_next(iter)) >= 0) {
		if (nodenum == master ||
		    nodenum == new_master)
			continue;

		ret = o2net_send_message(DLM_MIGRATE_REQUEST_MSG, dlm->key,
					 &migrate, sizeof(migrate), nodenum,
					 &status);
		if (ret < 0)
			mlog_errno(ret);
		else if (status < 0) {
			mlog(0, "migrate request (node %u) returned %d!\n",
			     nodenum, status);
			ret = status;
		}
	}

	if (ret < 0)
		mlog_errno(ret);

	mlog(0, "returning ret=%d\n", ret);
	return ret;
}


/* if there is an existing mle for this lockres, we now know who the master is.
 * (the one who sent us *this* message) we can clear it up right away.
 * since the process that put the mle on the list still has a reference to it,
 * we can unhash it now, set the master and wake the process.  as a result,
 * we will have no mle in the list to start with.  now we can add an mle for
 * the migration and this should be the only one found for those scanning the
 * list.  */
int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data)
{
	struct dlm_ctxt *dlm = data;
	struct dlm_lock_resource *res = NULL;
	struct dlm_migrate_request *migrate = (struct dlm_migrate_request *) msg->buf;
	struct dlm_master_list_entry *mle = NULL, *oldmle = NULL;
	const char *name;
	unsigned int namelen, hash;
	int ret = 0;

	if (!dlm_grab(dlm))
		return -EINVAL;

	name = migrate->name;
	namelen = migrate->namelen;
	hash = dlm_lockid_hash(name, namelen);

	/* preallocate.. if this fails, abort */
	mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache,
							 GFP_NOFS);

	if (!mle) {
		ret = -ENOMEM;
		goto leave;
	}

	/* check for pre-existing lock */
	spin_lock(&dlm->spinlock);
	res = __dlm_lookup_lockres(dlm, name, namelen, hash);
	spin_lock(&dlm->master_lock);

	if (res) {
		spin_lock(&res->spinlock);
		if (res->state & DLM_LOCK_RES_RECOVERING) {
			/* if all is working ok, this can only mean that we got
		 	* a migrate request from a node that we now see as
		 	* dead.  what can we do here?  drop it to the floor? */
			spin_unlock(&res->spinlock);
			mlog(ML_ERROR, "Got a migrate request, but the "
			     "lockres is marked as recovering!");
			kmem_cache_free(dlm_mle_cache, mle);
			ret = -EINVAL; /* need a better solution */
			goto unlock;
		}
		res->state |= DLM_LOCK_RES_MIGRATING;
		spin_unlock(&res->spinlock);
	}

	/* ignore status.  only nonzero status would BUG. */
	ret = dlm_add_migration_mle(dlm, res, mle, &oldmle,
				    name, namelen,
				    migrate->new_master,
				    migrate->master);

unlock:
	spin_unlock(&dlm->master_lock);
	spin_unlock(&dlm->spinlock);

	if (oldmle) {
		/* master is known, detach if not already detached */
		dlm_mle_detach_hb_events(dlm, oldmle);
		dlm_put_mle(oldmle);
	}

	if (res)
		dlm_lockres_put(res);
leave:
	dlm_put(dlm);
	return ret;
}

/* must be holding dlm->spinlock and dlm->master_lock
 * when adding a migration mle, we can clear any other mles
 * in the master list because we know with certainty that
 * the master is "master".  so we remove any old mle from
 * the list after setting it's master field, and then add
 * the new migration mle.  this way we can hold with the rule
 * of having only one mle for a given lock name at all times. */
static int dlm_add_migration_mle(struct dlm_ctxt *dlm,
				 struct dlm_lock_resource *res,
				 struct dlm_master_list_entry *mle,
				 struct dlm_master_list_entry **oldmle,
				 const char *name, unsigned int namelen,
				 u8 new_master, u8 master)
{
	int found;
	int ret = 0;

	*oldmle = NULL;

	mlog_entry_void();

	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&dlm->master_lock);

	/* caller is responsible for any ref taken here on oldmle */
	found = dlm_find_mle(dlm, oldmle, (char *)name, namelen);
	if (found) {
		struct dlm_master_list_entry *tmp = *oldmle;
		spin_lock(&tmp->spinlock);
		if (tmp->type == DLM_MLE_MIGRATION) {
			if (master == dlm->node_num) {
				/* ah another process raced me to it */
				mlog(0, "tried to migrate %.*s, but some "
				     "process beat me to it\n",
				     namelen, name);
				ret = -EEXIST;
			} else {
				/* bad.  2 NODES are trying to migrate! */
				mlog(ML_ERROR, "migration error  mle: "
				     "master=%u new_master=%u // request: "
				     "master=%u new_master=%u // "
				     "lockres=%.*s\n",
				     tmp->master, tmp->new_master,
				     master, new_master,
				     namelen, name);
				BUG();
			}
		} else {
			/* this is essentially what assert_master does */
			tmp->master = master;
			atomic_set(&tmp->woken, 1);
			wake_up(&tmp->wq);
			/* remove it from the list so that only one
			 * mle will be found */
			list_del_init(&tmp->list);
			__dlm_mle_detach_hb_events(dlm, mle);
		}
		spin_unlock(&tmp->spinlock);
	}

	/* now add a migration mle to the tail of the list */
	dlm_init_mle(mle, DLM_MLE_MIGRATION, dlm, res, name, namelen);
	mle->new_master = new_master;
	mle->master = master;
	/* do this for consistency with other mle types */
	set_bit(new_master, mle->maybe_map);
	list_add(&mle->list, &dlm->master_list);

	return ret;
}


void dlm_clean_master_list(struct dlm_ctxt *dlm, u8 dead_node)
{
	struct list_head *iter, *iter2;
	struct dlm_master_list_entry *mle;
	struct dlm_lock_resource *res;
	unsigned int hash;

	mlog_entry("dlm=%s, dead node=%u\n", dlm->name, dead_node);
top:
	assert_spin_locked(&dlm->spinlock);

	/* clean the master list */
	spin_lock(&dlm->master_lock);
	list_for_each_safe(iter, iter2, &dlm->master_list) {
		mle = list_entry(iter, struct dlm_master_list_entry, list);

		BUG_ON(mle->type != DLM_MLE_BLOCK &&
		       mle->type != DLM_MLE_MASTER &&
		       mle->type != DLM_MLE_MIGRATION);

		/* MASTER mles are initiated locally.  the waiting
		 * process will notice the node map change
		 * shortly.  let that happen as normal. */
		if (mle->type == DLM_MLE_MASTER)
			continue;


		/* BLOCK mles are initiated by other nodes.
		 * need to clean up if the dead node would have
		 * been the master. */
		if (mle->type == DLM_MLE_BLOCK) {
			int bit;

			spin_lock(&mle->spinlock);
			bit = find_next_bit(mle->maybe_map, O2NM_MAX_NODES, 0);
			if (bit != dead_node) {
				mlog(0, "mle found, but dead node %u would "
				     "not have been master\n", dead_node);
				spin_unlock(&mle->spinlock);
			} else {
				/* must drop the refcount by one since the
				 * assert_master will never arrive.  this
				 * may result in the mle being unlinked and
				 * freed, but there may still be a process
				 * waiting in the dlmlock path which is fine. */
				mlog(0, "node %u was expected master\n",
				     dead_node);
				atomic_set(&mle->woken, 1);
				spin_unlock(&mle->spinlock);
				wake_up(&mle->wq);
				/* do not need events any longer, so detach 
				 * from heartbeat */
				__dlm_mle_detach_hb_events(dlm, mle);
				__dlm_put_mle(mle);
			}
			continue;
		}

		/* everything else is a MIGRATION mle */

		/* the rule for MIGRATION mles is that the master
		 * becomes UNKNOWN if *either* the original or
		 * the new master dies.  all UNKNOWN lockreses
		 * are sent to whichever node becomes the recovery
		 * master.  the new master is responsible for
		 * determining if there is still a master for
		 * this lockres, or if he needs to take over
		 * mastery.  either way, this node should expect
		 * another message to resolve this. */
		if (mle->master != dead_node &&
		    mle->new_master != dead_node)
			continue;

		/* if we have reached this point, this mle needs to
		 * be removed from the list and freed. */

		/* remove from the list early.  NOTE: unlinking
		 * list_head while in list_for_each_safe */
		__dlm_mle_detach_hb_events(dlm, mle);
		spin_lock(&mle->spinlock);
		list_del_init(&mle->list);
		atomic_set(&mle->woken, 1);
		spin_unlock(&mle->spinlock);
		wake_up(&mle->wq);

		mlog(0, "%s: node %u died during migration from "
		     "%u to %u!\n", dlm->name, dead_node,
		     mle->master, mle->new_master);
		/* if there is a lockres associated with this
	 	 * mle, find it and set its owner to UNKNOWN */
		hash = dlm_lockid_hash(mle->u.name.name, mle->u.name.len);
		res = __dlm_lookup_lockres(dlm, mle->u.name.name,
					   mle->u.name.len, hash);
		if (res) {
			/* unfortunately if we hit this rare case, our
		 	 * lock ordering is messed.  we need to drop
		 	 * the master lock so that we can take the
		  	 * lockres lock, meaning that we will have to
			 * restart from the head of list. */
			spin_unlock(&dlm->master_lock);

			/* move lockres onto recovery list */
			spin_lock(&res->spinlock);
			dlm_set_lockres_owner(dlm, res,
				      	DLM_LOCK_RES_OWNER_UNKNOWN);
			dlm_move_lockres_to_recovery_list(dlm, res);
			spin_unlock(&res->spinlock);
			dlm_lockres_put(res);

			/* about to get rid of mle, detach from heartbeat */
			__dlm_mle_detach_hb_events(dlm, mle);

			/* dump the mle */
			spin_lock(&dlm->master_lock);
			__dlm_put_mle(mle);
			spin_unlock(&dlm->master_lock);

			/* restart */
			goto top;
		}

		/* this may be the last reference */
		__dlm_put_mle(mle);
	}
	spin_unlock(&dlm->master_lock);
}


int dlm_finish_migration(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
			 u8 old_master)
{
	struct dlm_node_iter iter;
	int ret = 0;

	spin_lock(&dlm->spinlock);
	dlm_node_iter_init(dlm->domain_map, &iter);
	clear_bit(old_master, iter.node_map);
	clear_bit(dlm->node_num, iter.node_map);
	spin_unlock(&dlm->spinlock);

	mlog(0, "now time to do a migrate request to other nodes\n");
	ret = dlm_do_migrate_request(dlm, res, old_master,
				     dlm->node_num, &iter);
	if (ret < 0) {
		mlog_errno(ret);
		goto leave;
	}

	mlog(0, "doing assert master of %.*s to all except the original node\n",
	     res->lockname.len, res->lockname.name);
	/* this call now finishes out the nodemap
	 * even if one or more nodes die */
	ret = dlm_do_assert_master(dlm, res->lockname.name,
				   res->lockname.len, iter.node_map,
				   DLM_ASSERT_MASTER_FINISH_MIGRATION);
	if (ret < 0) {
		/* no longer need to retry.  all living nodes contacted. */
		mlog_errno(ret);
		ret = 0;
	}

	memset(iter.node_map, 0, sizeof(iter.node_map));
	set_bit(old_master, iter.node_map);
	mlog(0, "doing assert master of %.*s back to %u\n",
	     res->lockname.len, res->lockname.name, old_master);
	ret = dlm_do_assert_master(dlm, res->lockname.name,
				   res->lockname.len, iter.node_map,
				   DLM_ASSERT_MASTER_FINISH_MIGRATION);
	if (ret < 0) {
		mlog(0, "assert master to original master failed "
		     "with %d.\n", ret);
		/* the only nonzero status here would be because of
		 * a dead original node.  we're done. */
		ret = 0;
	}

	/* all done, set the owner, clear the flag */
	spin_lock(&res->spinlock);
	dlm_set_lockres_owner(dlm, res, dlm->node_num);
	res->state &= ~DLM_LOCK_RES_MIGRATING;
	spin_unlock(&res->spinlock);
	/* re-dirty it on the new master */
	dlm_kick_thread(dlm, res);
	wake_up(&res->wq);
leave:
	return ret;
}

/*
 * LOCKRES AST REFCOUNT
 * this is integral to migration
 */

/* for future intent to call an ast, reserve one ahead of time.
 * this should be called only after waiting on the lockres
 * with dlm_wait_on_lockres, and while still holding the
 * spinlock after the call. */
void __dlm_lockres_reserve_ast(struct dlm_lock_resource *res)
{
	assert_spin_locked(&res->spinlock);
	if (res->state & DLM_LOCK_RES_MIGRATING) {
		__dlm_print_one_lock_resource(res);
	}
	BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);

	atomic_inc(&res->asts_reserved);
}

/*
 * used to drop the reserved ast, either because it went unused,
 * or because the ast/bast was actually called.
 *
 * also, if there is a pending migration on this lockres,
 * and this was the last pending ast on the lockres,
 * atomically set the MIGRATING flag before we drop the lock.
 * this is how we ensure that migration can proceed with no
 * asts in progress.  note that it is ok if the state of the
 * queues is such that a lock should be granted in the future
 * or that a bast should be fired, because the new master will
 * shuffle the lists on this lockres as soon as it is migrated.
 */
void dlm_lockres_release_ast(struct dlm_ctxt *dlm,
			     struct dlm_lock_resource *res)
{
	if (!atomic_dec_and_lock(&res->asts_reserved, &res->spinlock))
		return;

	if (!res->migration_pending) {
		spin_unlock(&res->spinlock);
		return;
	}

	BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
	res->migration_pending = 0;
	res->state |= DLM_LOCK_RES_MIGRATING;
	spin_unlock(&res->spinlock);
	wake_up(&res->wq);
	wake_up(&dlm->migration_wq);
}